The AI Floor ✳
A short game · one task · four drivers

One post,four drivers

The same request lands with four people at four levels of AI fluency. Watch what changes as the score goes up.

→ next beat   ← back
Before we start

Guess yourAI fluency score,out of 100.

Keep that number in your head.

The task

Promote the event

The requestHelp me write a social media post promoting our event.
How would you do it?
First idea that comes to mind. Shout it out.
Driver 1 of 4
34/100
The Copy-Paster
Developing tier

Uses AI regularly. Takes the output at face value.

34 · 55 · 70 · 85
TypesHelp me write a social media post promoting our event.
Then

Copies the first draft. Posts it.

2 MINUTES · ONE PROMPT · NO REVIEW

What it shows PromptingCritical Thinking A one-line instruction with no audience, no goal, no tone. The output is trusted as it comes. The AI decided everything, including what the event is for.
Driver 2 of 4
55/100
The Tactician
Developing tier

Gets things done with AI, fast and reliably.

34 · 55 · 70 · 85
TypesWrite a LinkedIn post announcing that our event is oversubscribed, with a call to action to sign up for future events. Write it the way Daniel Priestley would. Sincere, but clear in tone.
Then

Reads the draft carefully. Rejects a few lines. Edits the rest by hand before posting.

15 MINUTES · ONE GOOD PROMPT · A HUMAN EDIT

What it shows PromptingCritical Thinking The prompt carries the platform, the news, the call to action, a style reference and a tone. The output is read as a draft, not an answer. Judgement stays with the person.
Driver 3 of 4
70/100
The Conductor
Proficient tier

Orchestrates AI across the workflow, not just within it.

34 · 55 · 70 · 85
TypesI want to build a reusable skill for Claude called /post. I’ll give you a file of real, high-performing posts in this style and niche, a set of constraints, and a clear brief. Every time I run it, follow those.
Then
  1. Collects the examples. A markdown file of actual posts that performed, in this style, for this audience.
  2. Writes the constraints and the brief. Length, banned phrases, what a call to action must do, who the reader is.
  3. Saves it as a skill. From now on the request is one word, and the context comes with it every time.

AN HOUR ONCE · TWO MINUTES EVERY TIME AFTER · THE EXAMPLES DO THE TEACHING

What it shows PromptingWorkflowTechnical Understanding Context is built once and reused, not retyped. Real examples replace adjectives like “engaging”. The task has become a routine with a name.
Driver 4 of 4
85/100
The Architect
Advanced tier

Builds highly complex integrated systems using AI.

34 · 55 · 70 · 85
TypesBuild me a panel. Several sub-agents, each writing the post from a different angle. Here is what a good judge looks like and how the decision gets made. Generate twenty versions and rank them by confidence.
Then
  1. Reviews the winner against their own taste and judgement. Nothing is published until they think it is good enough. The panel proposes; the person decides.
  2. Works with the model they actually have. Its style, what it is good at, where it fails. Keeps an eye on what context it has been given and on the memory limits.
  3. If the task recurs, automates it. A scheduled job runs the panel and emails the winner. The person still signs off.

A SYSTEM, NOT A PROMPT · TWENTY CANDIDATES, ONE JUDGE, ONE HUMAN GATE

What it shows WorkflowCritical ThinkingTechnical UnderstandingPrompting Orchestration: many attempts, a defined judge, a ranked shortlist. Verification is built into the process, with a person as the last gate. Knows the model’s limits and the context it is working with.
Same request, four altitudes

What changes as the score goes up

34
The Copy-Paster
Asks, posts

One line in. First draft out. No review.

PromptingCritical Thinking
55
The Tactician
Briefs, edits

Platform, news, call to action, style, tone. Reads carefully, rejects, edits by hand.

PromptingCritical Thinking
70
The Conductor
Builds the skill

Real examples, constraints and a brief, saved once as /post. Two minutes every time after.

PromptingWorkflow
85
The Architect
Runs a panel

Sub-agents from different angles, a defined judge, twenty ranked versions, a human gate, then a scheduled job.

WorkflowCritical ThinkingTechnical

Same tools at every rung. What moves is how much of the thinking the person keeps, and how much of the doing they hand over.

1 / 8